lacti / object-detection-api

Very simple object detection from a photo with pretrained mscoco model.

Object detection api

Detect objects in a photo with pretrained ssd_mobilenet_V1_coco_11_06_2017 model.

External files

frozen_inference_graph.pb

1. Download ssd_mobilenet_v1_coco_11_06_2017.tar.gz
2. Unpack
3. Copy frozen_inference_graph.pb file to this directory from there.

mscoco_label_map.txt

Redefine mscoco_label_map.pbtxt as a list of (id, display_name).

Quick start

Install environment

If you want to use virtualenv like me,

python3 -m virtualenv venv
source venv/bin/activate

and then install other dependencies.

pip install -r requirements.txt

Inference

python inference.py image.jpg

Example

$ python inference.py image.jpg
Detect object from a photo: image.jpg
[('book', 0.3387419), ('bench', 0.11134028), ...]